Hacker News Reader

> Add on top of that neural atrophy, loss of oxygen, accumulation of senescent cells, poor clearance of waste products

I think the enjoyable thing about recreational drugs is the way they alter the normally-imperceptible boundaries between body/mind/Self. It provides an opportunity to understand what each of those things really are and what each contributes to my sense of being. I don't see these examples as all too different even if they are failure modes :)

So what did you understand that you did not know before?

Many people make the "opportunity to understand" claim, but I'm yet to hear anything meaningful coming out. Are you familiar with this joke: A guy decides to do drugs and keeps a notepad nearby to write down any profound revelations he has while high. The next morning he checks the notepad and finds he only wrote one thing: "The water is wet."

Here's this famous question (I can't readily find a source so I will say as I remember)

There is a scientist that studies color. She knows everything about it that can be possibly known. She knows the colors of everyday objects, she knows how colors work, how light's wavelength affects its transmission, etc etc. She is also fully colorblind (achromatopsia).

One day, she finally can see colors somehow. She isn't colorblind anymore.

The question is: now she can finally see the stuff she knew only intellectually, did she learn anything new?

Of course, she instantly sees color details of nearby objects down to a millimeter, so shit ton of new information. Bringing this back to the joke she can now look at a colored strip and write down the colors in the order they appear on it.

But the insinuation that drugs have similar effect is easily disproved by continued inability of the junkie to write down any new information.

Again, nobody dared to answer GGP question about examples of any revelations.

> Our conscious mind is much more of a facade than we typically imagine.

It patently is not, but I can see how it’s easy to believe it to be so. When you juxtapose the behavior of self-conscious organisms against reactive behavior, such as that observed during sentience-level consciousness, it’s easier to tease apart properties inherent to the conscious mind. For example, an LLM is going to have a hard time adjusting its behavior in an anticipatory manner to another organism, as their capacity for self-learning is currently limited.

But even relatively simple-minded animals with relatively higher-level cognition will adjust responses not only based on prompts, but also on their own nascent understanding of what’s expected behavior given a situation by continually adapting and forecasting, which is different from simply forecasting. I got interested in LLMs because of all the hype around their capabilities, but even simpler-than-human minds are doing much more complex stuff than generating the next probable step.

> Oliver Sacks' "The Man Who Mistook His Wife For A Hat"

Amazing book. Made me entirely reconsider how I think about human behaviour and "how they got like they did".

Should be required reading for being a human living in a society.

Sacks also wrote a book called "Hallucinations," which is excellent and highly relevant to this article. If any of you have loved ones who are starting to hallucinate or experience visual artifacts of the type associated with Charles Benoit Syndrome, I highly recommend reading this.

It's interesting to reflect upon who is "I" presented with these experiences. Brain might be hallucinating in a given moment due to infection or something. But the witness "I" is ever-intact experiencing come what may.

> When people say, "Well, LLMs are just generating token N+1 from the previous N tokens, they really aren't thinking", I counter with this: we have been having this discussion -- are you at all aware of the stream of words coming out of your mouth, or are you hearing them the same time that I am?

Neither awareness of the stream of words coming out of my mouth, nor hearing them simultaneous with any audience proves or disproves LLM's generating token N+1 from previous N tokens.

I believe we both disagree with the irrelevant observation that "Well, LLMs are just generating token N+1 from the previous N tokens, they really aren't thinking"

But I also disagree with the idea that we are strictly passively aware of what we express: that would constitute a return to fundamental dualism (separation of body and mind), conventionally called dualism.

Dualism is a pragmatic approximation to make to model peoples thoughts and behaviors, but there is a reason dualism was superseded by materialism (the philosophical standpoint, not talking about materialistic ideology), consider the following thought experiment: suppose one believes in a type of fundamental dualism (not just using dualist approximations for computational efficiency considerations) where the sentient "mind" is passively experiencing what happens in the material brain, experiencing emotions, colors, tastes, smells, concepts, thoughts, ... then wouldn't it be such a coincidence that this materialistic brain happens to describe precisely what you seem to experience? a fundamental theory of dualism where the mind can not influence the real world -which includes the brain- but where the real world can influence the brain and the mind is utterly unlikely: if the mind could not affect the operation of the brain, how can the subjective experience which the mind perceives ever end up in the material brain so it can be materially pronounced etc?

For example, a human brain is composed of cells (neurons), a nation is composed of human citizens.

Suppose you make the claim:

"My subjective experience is neither that of a single neuron, surrounded by other neurons; nor that of a nation, surrounded by other nations in a geopolitic landscape; no my subjective experience is at the level of a human with its single brain"

Then I believe 99% of the population would agree with the above statement. But if simultaneously the dualist worldview were correct:

* that the material world obeys some rules of physics (known and unknown, indeterministic or not),

* that the subjective experience is a oneway projection of what happens to a limited subset of the real world,

* that these distinct "minds" passively observe what happens in the brain, and this indirectly, what happens in our sensory organs, and this indirectly what happens in the outside world;

* that these "minds" can exert no force or influence on the brain, and thus not on the muscles, and thus not on the outside world

... then it would be an extremely unlikely coincidence that human brains and bodies are expressing these facts which happen to correspond to our supposedly 100% passive observing "minds".

Statistics and likelihoods are part of mathematics, and indifferent to physical postulates. Mathematics is part of logic, indifferent to physical postulates. The passive dualist hypothesis can be rejected with arbitrarily high confidence, just ask more and more people if their subjective experience is closer to:

a) one neuron among many b) one human among many c) one nation among many

So we know that this form of dualism is subjectively falsifiable, just like others can not prove their experience matches b) more than a) and c); neither can I prove that my experience matches b) more than a) or c). But subjectively this is being proven to me all the time, I don't need to see proof of b) I experience it every day, all life long.

So IF dualism does nevertheless hold, I have proven to myself it can not be this oneway passive delocalized homunculus form of dualism.

So we have proven that IF dualism holds, that this supposed "mind" MUST be able to perturb the material body's "brain" and thence muscle, and outside world.

But at this point we are basically saying that not just the brain but the "mind" must be part of the physical universe, since it is interacting both ways with the material brain. No distinction between physical variables of the brain and physical variables are presented, apart from some insinuated specialness of "mind" quantities (variables in physics) and "non-mind" quantities (say a neuronal spike), without any definition how to even recognize the difference.

If the "mind" quantities can influence the "non-mind" quantities and vice versa, then we should be able to device measurement apparatuses to measure not just the "non-mind" quantities in the brain (like EEG signals, etc), but also the "mind" quantities (to which the "non-mind" signals like EEG supposedly merely spuriously correlate). Without any specification of their supposed specialness compared to the usual explainable-by-boring-physics-probabilities, the goalposts can always be moved: perhaps its little angels feeling the neuronal spikes and those same little angels slightly delaying or enhancing/quenching probabilities of those neuronal spikes: foreseeably every newly discovered deviation of Nth order approximation, together with its corresponding Neuther theorem quasiparticle, yadda yadda yadda, will be initially hailed as vindication of dualism ("so there is possibly a separate Mind!"), and subsequently invalidated after arduous reconsiderations by the materialists and reductionists, and the deviations will turn out to be explained by considering the N+1th order approximations of the rules of physics...

So we discover that this second type of dualism (with both way interaction) is unfalsifiable, and only superfluous, and hence not the domain of science, but mysticism. Might as well say that invisible angels are pushing the Moon in an orbit around the earth, and pushing the Earth in an orbit around the sun, etc... until one is basically saying the invisible angels are executing physics (probabilistic or not) to the T, time and time again. So basically the angels contribute nothing to the predictions of physics.

So dualism was rejected in favor of materialism, even precisely because everybody answers b); not despite it!

Not the person you replied to, but I totally agreed with him and, interestingly, cannot make sense at all of your reply. Those points especially seem to be the crux of your argument but I don't see the logic supporting them:

> then wouldn't it be such a coincidence that this materialistic brain happens to describe precisely what you seem to experience?

> then it would be an extremely unlikely coincidence that human brains and bodies are expressing these facts which happen to correspond to our supposedly 100% passive observing "minds".

I fail to see what's unlikely about the viewpoint that the mind is passively watching the movie that our body is the main character of; Or why such a one-way relationship would contradict the mind being real and "physical". Physics is full of one way functions. It is not the first time that I totally fail to comprehend one side of this very debate, and it feels like I'm missing an essential insight.

I understand that in dualism the "mind" (the conscious experience) does not bring anything to the physical world and look therefore utterly useless, which is not satisfactory but does not prove or disprove anything (that's what I understand to be "the hard problem of science"), like the angels pushing the celestial bodies according to the laws of physics. I would gladly rule out the possibility of such useless conscious experience like I rule out the theory of the little angels, if not for the personal experience I have of it :)

So, to follow along that analogy, it seems to me that yes indeed the moon obeys the law of physics but still somehow some little angels that I can see pretend that they push it around. And your argument sounds like "there must be an influence from these angels to the laws of physics because otherwise it's too unlikely that those angels would be seen", which sounds suspicious to me given how the laws of physics stands very well on their own.

I don't know if I've made the discussion clearer or more opaque. Would you be able to rephrase your explanation differently?

Consider the following visual phenomena / optical illusion:

https://michaelbach.de/ot/sze-silhouette/index.html

The silhoute may be perceived as rotating clockwise or counterclockwise.

There is not in fact enough information to determine which direction objectively.

Now suppose 2 types of experiments:

A) someone wears spy glasses and observes this effect, and is tasked to regularly speak out what direction (s)he perceives the figurine to spin. The spy glasses records what the subject sees, but also these remarks.

B) you observe this animation yourself, not wearing spy glasses, and you yourself speak out loud what direction you see the figurine spin, including occasional flips (which occur more frequently on faster / smaller variants of such animations)...

... now ...

Suppose you are watching a recording of someone else's A) variant, effectively you may be seen as a passive spectator of a prerecording of this other person. While watching this spy glass prerecording you would have your own interpretation (clockwise or not) and occasional flips of direction, at the same time you would be hearing this other person's interpretation, and you might agree with this other person half the time and not for half the other, the timing of the flips wouldn't correspond etc. Even guessing a coin toss will be correct half the time.

However if you yourself perform B) and speak for yourself, the statistics change, and the rotation direction your supposed 100% passive "mind" perceives corresponds with what your material body is speaking out loud all the time.

How do you explain that discrepancy in statistics? How was your brain and body able to systematically express what your supposedly 100% passive "mind" homunculus observed?

Hence you have "proven" your "passive mind spectator" to actually be active, and thus necessarily BE MEASURABLE. You come to the conclusion you are a materialist after all, even though the dualist simplifaction of the world is useful and pragmatic, you have shown it to be fundamentally wrong.

I wrote "proven" in quotes because you can't prove to a third person this correspondence between (counter)clockwise perception and verbalization.

But you have certainly proven this to yourself.

I'm going to be naive and open about how your answer make me feel because I believe it may help understand why and where we don't speak the same language, at the risk of giving the impression that I'm looking down at it. Be assured that it is not the case, please bear with me.

I don't see the point of the "spy glasses" since the animation we see with or without the glasses is exactly the same - supposedly the glasses do not alter the vision so the flips experienced by the observer would be the same with or without the glasses.

> Suppose you are watching a recording of someone else's A) variant, effectively you may be seen as a passive spectator of a prerecording of this other person

I assume you mean I'm watching at the recording made by the glasses, so that I see kind of "through the eyes" of the previous observer.

I do not fully understand why you had to resort to a visual illusion. If the participant is just asked to say a random number between 1 and 10 every minute, and I'm listening to the recording of the experiment while at the same time being subjected to it myself, I will also utter the same number once every 10 minutes in average. And of course, in the "B" version of that variant of the experiment based on random numbers, I would "agree" with me all the time because there is nothing else to compare to.

I don't see the connection with the question of

> How do you explain that discrepancy in statistics?

This sounds silly. I explain the different result from the fact that it's a different experiment. In the first case I'm comparing my own "random flips" to someone elses, and in the second I'm doing no such comparison.

> How was your brain and body able to systematically express what your supposedly 100% passive "mind" homunculus observed?

My passive "homunculus" does not observe the movie of the rotating figure, and does no say anything about how it experiences the movie. It merely observes my brain interpretation of the rotating figure, while it listens to my brain uttering in which direction the figure is rotating. It is "out of the loop".

(for the purpose of this discussion, the brain is just part of the body, right?)

Using the same experiment, let me try to demonstrate that the experiencing subject can only be passive.

I do not know how you stand on the "neural networks do/do not experience self consciousness" debate, and it does not matter. For this though experiment all I need is that you agree that it is possible in theory to build a large neural network that experiences _no_ consciousness whatsoever. Just a stupid machine made of cogs and gears if you prefer that to a weight matrix, but enough of them that it has the capacity to be given a set of pictures as input and, if that set of pictures represent a movie of some rotating object, it would flash a green light if the object is rotating clockwise, and a red one if it's rotating counter-clockwise.

Of course, if subjected to the optical illusion of the rotating ballerina, the machine would be confused and, sometime flashes red and sometimes green.

Now, in principle, we can improve this machine to make it more "brain-like" by enlarging the simulation (or adding more cogs), until eventually we can simulate the exact body (including the brain) of the human being participating in your experiment well enough that we can predict in which direction he will see the figure rotating and when he will experience flips, because the machine will see the exact same thing in the same way and, reproducing the same brain circuitry than the human subject, will be confused in exactly the same way at the same moment.

So, we can predict what the subject will "experience" with a machine that does not experience anything. Therefore, the fact that the human subject experiences anything at all is totally irrelevant to what he will say the rotation is (like the angels "pushing" the moon are irrelevant to explain the motion of the moon).

And actually, similar experiences are being done at least since the 90s ; not with a large brain simulation but just by measuring the brain activity of the subject and being able to predict what he is going to say before he actually has a conscious experience of it. Three papers from a quick internet search:

- https://www.nature.com/articles/news.2008.751

- https://www.pnas.org/doi/abs/10.1073/pnas.1212218110

- https://www.nature.com/articles/s41598-019-39813-y

Don't you think that all this point strongly toward a "one way function"?

Thanks for the extensive response. However, I'm a materialist and I don't follow your point. You say dualism was replaced by materialism, but there are still billions who subscribe to dualism; it wasn't replaced.

My claim isn't that the current LLMs are thinking; my claim is that much of what comes out of our mouths aren't any more sophisticated.

Speak for yourself

I just reread my response. "I'm a materialist", "I don't follow your point", and "my claim". Whose mouth do you think I'm putting words into?

The one statement I made that wasn't just about me was "billions subscribe to dualism" -- which doesn't speak for any specific person. Do you disagree that dualism is still very much alive?

Modern materialism (since the 17th century) defines the material world by that which is quantifiable and hence measurable. Consciousness is not wholly characterised by what we can measure. Therefore, if one believes that consciousness exists, then necessarily materialism is false.

This widespread notion that whatever affects the material world is itself material fails to understand what both materialism and dualism means.